Brainforge Slack Assistant — Discovery: more ways to use it

Date: 2026-03-26 (updated 2026-03-27)
Primary intent: Identify additional ways the internal Brainforge Assistant can be used (new habits, channels, workflows, and prompts)—not only reliability or ops hygiene.
Primary evidence (target): Snowflake RAW — full Slack message history (see snowflake-slack-assistant-discovery.sql).
Supplementary: Slack MCP keyword search (noisy); useful for anecdotes only.


Executive summary

Slack search alone under-counts conversational use, but it still surfaces patterns worth doubling down on and gaps worth filling. The assistant is already positioned as GTM research (e.g. company sizing with live web), validation / thread-QA (e.g. repeat metrics from thread context), ops routing (/request-access, /request-expense), and repo-aware answers (/brainforge code, platform context). The main opportunity is to name and promote a broader menu of uses—meeting follow-ups, client-channel internal asks, delivery lead briefs, onboarding shortcuts, and “ask the vault” behaviors from ROADMAP.md—so adoption is intentional rather than accidental.

Recommended focus: (1) Publish a short “Ways to use @brainforge” internal playbook (channels + example prompts). (2) Pilot 2–3 new use cases per quarter with one champion each (e.g. Marketing: account research; Delivery: thread recap before client sync; CSO: prep from last Slack + Linear). (3) Treat tooling fixes (live lookup PLT-1233, slash-command ownership, logging) as enablers, not the headline.


Goals and non-goals

Goals

  • Expand documented and habitual uses of @brainforge / /brainforge beyond ad-hoc chat.
  • Tie Slack signals and roadmap items to concrete scenarios (“when X happens, ask Y”).
  • Make it easy for new hires to copy example prompts per team.

Non-goals

  • Replace client-specific bots or n8n workflows; those are separate products.
  • Use keyword Slack search as the only measure of success.

Methods

A. Snowflake (primary — run locally)

StepAction
1Connect with your user (1Password Employee vault: “Brainforge Internal Snowflake”) via Snowflake UI, /opt/homebrew/bin/snow, or Python. Use account KOWHRBS-TS55708 (or locator form from your vault URL).
2Run SHOW TABLES in RAW.DLT_DLT_SLACK (and staging / SLACK_DB* if needed) — see move-dbs-into-raw.sql.
3DESCRIBE the message fact table; map text / user / channel / ts columns.
4Filter last 90 days for assistant-adjacent traffic: @brainforge, /brainforge, U08CTJAJYP5 (bot), request-access, request-expense. Export a 500-row sample for labeling.
5Apply the taxonomy below; aggregate counts by category → prioritize roadmap.

Agent environment note (2026-03-27): Automated Python connection from this workspace failed (Could not connect to Snowflake backend / retryable login). Treat Snowflake analysis as local-only until network/auth is reliable from CI or your laptop.

SQL starter: snowflake-slack-assistant-discovery.sql

B. Slack MCP (supplementary)

InputLimitation
slack_search_public_and_private with after:2025-12-26Keyword-only; same bot user posts Zoom webhooks and integrations—not clean usage stats

C. Other signals

Structured app logs (slash vs app_mention vs webhook), short user survey, #ai-test-channel prompt library.


Taxonomy (label each sampled message)

Use one primary category per row; add secondary tag if useful.

CodeCategoryDefinition
GTMAccount / market researchURLs, sizing, competitors, metrics questions
ENGRepo / platform / /brainforge codebrainforge-platform, PRs, infra, debugging
OPSAccess / expense / process/request-access, /request-expense, Linear Asks refs
DELDelivery / client contextInternal #client-* asks, timelines, blockers
THREADThread-only / recap“from this thread only,” summarize, repeat metrics
METANoise / integrationZoom webhooks, empty bot posts, non-human pipelines
OTHERUncategorizedNeeds follow-up

Volume KPIs (fill after Snowflake run): % of rows in META (target: separate routing so META ≠ “usage”); top 3 non-META categories; week-over-week trend optional.


Prioritized feature roadmap (from taxonomy + ROADMAP)

Rank with RICE or simple Impact × Feasibility. Replace (TBD) after Snowflake counts land.

PriorityThemeFeatureSupports taxonomyDepends on
P0GTMReliable live web / company metrics in-threadGTMPLT-1233
P0AdoptionPlaybook + example prompts by categoryAllDocs
P1KnowledgeVault RAG for “how we work”ENG, OPS, DELROADMAP / infra
P1DeliveryPer-channel or client-scoped contextDELROADMAP
P2ThreadStrong thread context + no re-ask for pasted contentTHREADassistant + tests
P2OpsUnified /request-* UX clarityOPSSlack app
P3AmbientNudges, approval-before-postMETA reductionROADMAP
EnablersSlash collision fix; structured loggingMeasurementApp + Slack admin

More ways to use Brainforge Assistant (core deliverable)

This section answers: what should we encourage people to do with it? Evidence tags: [Slack] from MCP search, [Roadmap] from internal doc, [Product] from app capabilities.

A. Already evidenced or implied [Slack] / [Product]

UseWhat it looks likeWhy it matters
Account & market sizing[Slack] @brainforge in marketing with a URL; user wants traffic, social, revenue estimates → surfaced need for live lookups (PLT-1233).Reusable GTM pattern: “size this prospect” before calls.
Thread-grounded answers[Slack] Ask to repeat or consolidate items only from the thread (e.g. metrics list)—assistant complied in one shot.Reduces re-pasting; good for PLT/validation and long threads.
Repo / platform questions[Product] /brainforge code … and repo-context API—answer with links to brainforge-platform.Default habit for engineers and CSOs touching the monorepo.
Ops self-serve[Slack] /request-access, /request-expense + Linear Asks “General Ops Request.”Keeps Ops out of DMs; assistant as front door.

B. High-value additions [Roadmap] / [Product] (under-promoted today)

UseExample prompt or triggerDepends on
Vault-grounded policy / process“What’s our commit format?” “Where is the SOW framework?”Stronger vault RAG wiring [Roadmap]; today partial via repo context.
Meeting → actions“From the last Zoom in this channel, what are the open actions?”Zoom notifications already land in ai-test-channel; next step is summarization + ticket suggestions [Roadmap].
Client-internal channel copilotIn #client-* internal channels: “Summarize blocker threads this week” / “What did we promise?”Per-channel context [Roadmap]; needs scope guardrails.
Delivery lead brief“Pull top risks from this thread + link Linear tickets mentioned.”Slack thread read + Linear MCP-style context (manual or automated).
Onboarding“How do I request tool X?” “Where is the Data Platform doc template?”Combine standards links + ops commands.

C. Creative / stretch (vision from [Roadmap])

  • Ambient nudges (unanswered threads, hours reminders)—assistant as nudge, not chat-only.
  • Approval-before-post for sensitive replies—human clicks Approve in Slack.
  • Per-team shortcuts (e.g. sales vs engineering different default context).

Slack search: what it did and did not prove

  • Did show: Marketing research + thread QA are real internal patterns; Zoom firehose dominates ai-test-channel (integration noise in search).
  • Did not show: Volume of true @brainforge Q&A (keyword search mixes GitHub URLs, “brainforge” in Notion links, client feedback pipelines).

Use this section only to support use-case ideas, not to rank them statistically.


Prioritized workstreams (reordered for “more usage”)

Phase 1 — Adoption and clarity (lead)

  1. Internal playbook — One-pager: “Ways to use Brainforge Assistant” (channels, @brainforge vs /brainforge, 5–10 example prompts per function: GTM, Eng, Ops, Delivery).
  2. Champion pilots — Pick 2 pilots (e.g. Marketing: weekly account prep; Eng: repo question Fridays) for 4 weeks; collect 3 bullet retrospective each.

Phase 2 — Capabilities that unlock new uses

  1. Live web / metrics (PLT-1233) — Unlocks marketing & partnership research as promoted in the playbook.
  2. Vault RAG [Roadmap] — Unlocks “ask Brainforge how we work” without hunting Notion.

Phase 3 — Enablers (supporting)

  1. Slash-command ownership — Fix test vs prod /brainforge collision so documented commands match reality.
  2. Structured logging — Separate conversational use from webhooks for honest adoption metrics.
  3. Ops UX parity — Clear copy for /request-access vs /request-expense so ops stays a trusted surface.

Risks and dependencies

  • Discovery vs noise: Same Slack app identity for integrations obscures “real” chat usage until logging exists.
  • Expectations: Promoting “live research” requires PLT-1233 (or equivalent) to land, or users churn.

Success metrics (oriented to adoption)

MetricTarget
Playbook publishedInternal link shared in brainforge-team or equivalent
Pilot completion2 pilots run; 5+ example prompts added to playbook from real usage
Reported helpfulnessOptional emoji feedback on bot replies [Roadmap]
PLT-1233Closed; marketing re-test

Appendix A — Linear-ready tickets (use-case + adoption first)

OrderSuggested titleTypeNotes
0Run Snowflake discovery SQL on RAW Slack tables; label 500-row sample with taxonomy; fill volume KPIschoreUses snowflake-slack-assistant-discovery.sql; blocks data-driven priority tweaks
1Publish internal “Ways to use @brainforge / /brainforge” playbook with example prompts by teamdocsPrimary adoption deliverable
2Run two 4-week use-case pilots (e.g. Marketing account research + Eng repo Q&A) and capture prompts for playbookchoreDepends on 1 minimally; can run in parallel
3Verify and close PLT-1233 (live company-metrics / lookup flow for marketing-style asks)fix / QAUnlocks promoted GTM use case
4Spike: vault-grounded answers in Slack for top 10 “how we work” questionsfeatRoadmap RAG; scoped list
5Fix /brainforge slash command collision between prod and test Slack appsfixEnabler; not a “use case” by itself
6Add structured logging (slash vs mention vs webhook) for adoption metricsfeatEnabler

Already tracked: PLT-1233.


Appendix B — References